Research On Covering Rough Set Theory And Catalogue Search

Rough set theory is one of the important tools for data mining, and is also a branch of granular computing. The classical rough set theory is proposed by Poland Pawlak in 1980s. Rough set theory provides a set of systemic methods for dealing with the incomplete and inadequacy information by undistinguishable relation (equivalence relation). However, the application range of equivalence relation is finite. With extending the information partition which is produced by equivalence relation to information covering, Zakowski et al. expanded the Pawlak rough set theory to covering generalized rough set theory. Recently, many scholars had put forward several covering rough set models, and also discussed some related properties and their interrelationship.Based on the existed covering generalized theory, this thesis extends the covering rough set theory and applies it to the catalogue search. The main works are:(1) In this thesis, it firstly summarizes several covering rough set models and then concludes the fact that the main differences of these models are the distinctions of the upper approximation as well as the lower approximation.(2) According to the fact that there exists redundant situation after the reduction of the covering space, this thesis proposes the concept of the including reduction.(3) On the basis of the definition of information entropy about partition space which is defined by scholars, this thesis develops the rough information entropy which is suitable for covering space and the information expression schemes for both of the reduction and the including reduction.(4) By combining with the structural fact of the covering space, this thesis defines structural approximation operators and multi-level covering space.(5) Based on the multi-level covering space, this thesis gives the concept of catalogue tree, and also discusses its reduction and approximation computing.(6) By combining with the features of the catalogue search system, this thesis realizes the production, updating and related query options of the catalogue tree. Moreover, this thesis proposes some new query operators and evaluation of the closeness degree.